DraftDraft 1 Safety Evaluation of Unconventional Outside Left-Turn Lane Using Automated Traffic...

Draft

Safety Evaluation of Unconventional Outside Left-Turn Lane

Using Automated Traffic Conflict Techniques

Journal: Canadian Journal of Civil Engineering

Manuscript ID cjce-2015-0478.R1

Manuscript Type: Article

Date Submitted by the Author: 07-Feb-2016

Complete List of Authors: Guo, Yanyong; Southeast University Sayed, Tarek; Department of Civil Engineering, Zaki, Mohamed; University of British Columbia, Civil Engineering Liu, Pan; Southeast Univeristy, School of transportation

Keyword: safety analysis, traffic conflicts, computer vision, outside left-turn lanes

https://mc06.manuscriptcentral.com/cjce-pubs

Canadian Journal of Civil Engineering

Draft

1

Safety Evaluation of Unconventional Outside Left-Turn Lane Using Automated Traffic

Conflict Techniques

By

Yanyong Guo

Jiangsu Key Laboratory of Urban ITS

Jiangsu Collaborative Innovation Center of Modern Urban Traffic Technologies

Southeast University

Si Pai Lou #2, Nanjing, China, 210096

Email: [email protected]

Tarek Sayed

Department of Civil Engineering, University of British Columbia

6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4


Mohamed H. Zaki

Department of Civil Engineering, University of British Columbia

6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4

Email:[email protected]

Pan Liu (Corresponding author)

Jiangsu Key Laboratory of Urban ITS

Jiangsu Collaborative Innovation Center of Modern Urban Traffic Technologies

Southeast University

Si Pai Lou #2, Nanjing, China, 210096

Tel: (+86-025)83791816


Submitted to Canada Journal of Civil Engineering

Page 1 of 42



Draft

2

Abstract: The objective of this study is to evaluate the safety impacts of unconventional outside

left-turn lane at signalized intersections. New designed unconventional outside left-turn lane are

increasingly used at signalized intersections in urban areas in China. The unconventional outside

left-turn lane design allows an exclusive left-turn lane to be located to the right of through lanes

to improve the efficiency and increase the capacity of left-turn movements. However, the design

also raises some concerns regarding potential negative safety impacts. The evaluation is

conducted using an automated video-based traffic conflict technique. The traffic conflicts

approach provides better understanding of collision contributing factors and the failure

mechanism that leads to road collisions. Traffic conflicts are automatically detected and time to

collision (TTC) is calculated based on the analysis of the vehicles’ positions in space and time.

Video data is collected from a signalized intersection in Nanjing, China, where both traditional

inside and unconventional outside left-turn lanes are installed on two intersection approaches.

The other two approaches have only inside left-turn lanes. The study compared frequency and

severity of conflict for left-turning vehicles as well as the percentage of vehicles involved in

conflicts from the inside and outside left-turn lanes. The results show that the intersection

approaches with outside left-turn lanes had considerably more conflicts compared to approaches

without outside left-turn lanes. As well, the approaches with outside left-turn lanes had

significantly higher conflict severity than the approaches without outside left-turn lanes. As such,

it is recommended that the trade-off between the improved mobility and negative safety impact of

outside left-turn lanes be carefully considered before recommending their installation.

Keywords: safety analysis; traffic conflicts; computer vision; outside left-turn lanes

Page 2 of 42



Draft

3

1. Introduction

China has seen significant increase in private vehicle ownership over the past two decades which

has resulted in high levels of traffic congestion. In many large cities in China, transportation

engineers are being challenged by the ever-increasing traffic demand and the corresponding

congestion at at-grade signalized intersections. One of the most important factors that impact the

performance of at-grade signalized intersections is the presence of heavy left-turn movements

(Tabernero and Sayed 2006). Therefore, several measures to improve the performance of

intersections with heavy left-turn movements have been investigated, some of which have been

unconventional schemes. Traditional measures to improve the performance of intersections with

heavy left-turn movements include signal timing optimization, adding exclusive left-turn lanes,

and signal coordination. However, in many cities, these traditional measures have been exhausted

and are no longer able to alleviate congestion at signalized intersections (Hummer 1998). Other

solutions such as grade separation are costly and aesthetically unpleasing (Goldblatt et al. 1994).

As such, the use of unconventional intersection designs has been advocated as a solution to

improve traffic operations at at-grade signalized intersections (Autey and Sayed 2013; El Esawey

and Sayed 2013). Several unconventional measures for treating left-turns have been developed

including the Median U-turn, the Crossover Displaced Left-Turn (XDL), and the Upstream

Signalized Crossover (USC) schemes (Liu et al. 2007; El Esawey and Sayed 2013). In China, a

new design, the unconventional outside left-turn lane, has been progressively used at signalized

intersections in many urban areas. The unconventional outside left-turn lane design allows an

exclusive left-turn lane to be located to the right of through lanes, which improves the efficiency

of left-turn vehicles (Figure 1). Protected left-turn phases are also provided to accommodate the

left-turn movements at these signalized intersections.

Page 3 of 42



Draft

4

The unconventional outside left-turn lane design is often used on major urban arterials in

several Chinese cities under the following conditions: (a) when there is a bus station located

within the upstream function area of a signalized intersection and a large number of buses need to

make left turns at the intersection; (b) when a major right-side street is located close to an

intersection and a high number of vehicles from the driveway need to make left turns at the

intersection; and (c) when an urban interchange ramp terminal is located within the upstream

functional area of a signalized intersection and a large number of vehicles from the ramp terminal

need to make left turns at the intersection.

There are two main advantages associated with the outside left-turn lane: (a) it provides

larger left-turn radius for left-turn movements of heavy vehicles and buses, and (b) left-turning

vehicles coming from upstream driveways or bus stations do not have to make lane changes to

the most inside lane to make left turns. As such, it is supposed to simplify driving tasks and to

reduce conflict points on approaches to signalized intersections. However, the use of

unconventional outside left-turn lane has also raised some safety issues. There are two main

shortcomings: (a) vehicles in through traffic may mistakenly enter the outside left-turn lane and

will cause safety problems at signalized intersections when they are trying to merge back to

through traffic; and (b) the presence of two left-turn traffic streams at one approach may result in

high number of conflicts inside the signalized intersection.

Therefore, there is an important need to conduct a safety evaluation for this

unconventional design. Traditionally, safety evaluations were mainly based on collision data.

However, reliable collision data may not always be available in many jurisdictions. This is

particularly true for China, where the reliability and availability of crash data are often

problematic. Additionally, an adequate number of collisions are required to occur and be

recorded over an adequately long period in order to conduct a sound safety evaluation. The use of

Page 4 of 42



Draft

5

traffic conflicts for safety evaluation has been gaining acceptance as a surrogate for collision data

analysis (Amundsen and Hydén 1977; Sayed and Zein 1999; Tarko et al. 2009). The traffic

conflicts approach provides better understanding of collision contributing factors and the failure

mechanism that leads to road collisions. As well, traffic conflicts are more frequent than road

collisions and are of marginal social cost.

The objective of this study is to evaluate the safety impacts of unconventional outside

left-turn lane at signalized intersections. The evaluation was conducted using a video-based

automated traffic conflict technique. Video data were collected from a signalized intersection in

Nanjing, China, where both traditional inside and unconventional outside left-turn lanes are

present. The study compared the severity and frequency of conflict for left-turning vehicles from

inside and outside left-turn lanes.

2. Previous Work

2.1 Unconventional outside left-turn lane

Although the use of outside left-turn lanes has been recently increasing in larger cities in China,

little information and documentation of their use are available about this unconventional design.

A review of the literatures regarding the outside left-turn lane design found only a few studies

related to the operation evaluation and drivers’ selection of the unconventional outside left-turn

lanes. No studies were found evaluating the safety of outside left-turn lanes. A recent study by

Liu et al. (Liu et al. 2011) compared the left-turning speed and saturation flow rate from the

outside left-turn lane and the inside left-turn lane. It was reported that the outside left-turn lane

provided a significantly higher speed of passenger cars and heavy vehicles than that of the inside

left-turn lane. However, the higher speeds did not result in higher left-turn lane capacity. It was

Page 5 of 42



Draft

6

found that the left-turn saturation flow rate of the outside left-turn lane was slightly lower, but

generally comparable to that of the inside left-turn lane.

In a subsequent study conducted by Liu et al. (Liu et al. 2013), the factors affecting

driver’s selection of the outside left-turn lanes at the signalized intersections were identified using

a binary logit model. The results showed that the likelihood of drivers choosing the outside left-

turn lane would increase with an increase in traffic volume on the major road and the queue

length in the inside left-turn lane; while it would decrease with an increase in the distance from

the upstream driveway to the signalized intersection. Furthermore, heavy vehicle drivers and

those from the upstream driveway were more likely to select the outside left-turn lanes. Liu et al.

(Liu et al. 2013) also developed a procedure to determine if outside left-turn lanes were needed

given traffic conditions and geometric characteristics of a signalized intersection.

2.2 Conflict analysis

The traffic conflict technique (TCT) is probably the most developed surrogate safety assessment

method. The objective of TCT is to identify the frequency and severity of near misses which

enables the assessment of the safety performance of a road facility without waiting for collisions

to occur. The traffic conflict technique was first developed by Perkins and Harris (Perkins and

Harris 1968). The definition of traffic conflicts has since been revised with an accepted definition

proposed by Amundsen and Hydén (Amundsen and Hydén 1977) as “an observable situation in

which two or more road users approach each other in space and time to such an extent that there

is a risk of collision if their movements remain unchanged”.

Various conflict indicators have been developed to measure the severity of a traffic

conflict by quantifying the spatial and temporal proximity of road users involved in the

interaction (Brown 1994; Tarko et al. 2009). The main advantage of conflict indicators is their

ability to capture the severity of an interaction in an objective and quantitative way (Autey et al.

Page 6 of 42



Draft

7

2012). A commonly used traffic conflict indicator is the time-to-collision (TTC) defined as

“…the time that remains until a collision between two vehicles would have occurred if the

collision course and speed difference are maintained” (Hayward 1972).

The traffic conflict technique involves the observation, recording, and evaluation of the

frequency and severity of traffic conflicts at a location. The observation was traditionally

conducted manually by a team of trained observers. However, manual conflict observation has

been challenged by several issues, such as cost of training observers and reliability and

consistency of the result (Sayed et al. 1994). Recently, video-based automated conflicts analysis

was advocated as an alternative conflicts data collection procedure (Saunier et al. 2010). The

automated approach provides a more practical and efficient way to capture, store and analyze

elements of the traffic information. It provides a guideline for the objective measurements of

conflicts indicators (Laureshyn et al. 2009; Sayed et al. 2012). As well, it overcomes the

shortcomings of field-based methods which are unreliable, time consuming and costly.

Furthermore, video data represents a permanent record of the traffic events analyzed and can be

reviewed and validated. The application of video-based automated conflict analysis was

demonstrated to be very useful in conducting various safety analysis such as before-after safety

evaluation (Autey et al. 2012; Tageldin et al. 2014), bicycle related data collection and safety

diagnosis (Zaki et al. 2013; Sayed et al. 2013), and pedestrian related safety analysis (Ismail et al.

2010; Zaki et al. 2013).

3. Study Location

Video data were collected from a signalized intersection in Nanjing, China. Nanjing is the capital

city of the Jiangsu province in the south area of China. The number of motor vehicles in Nanjing

reached two million by the end of 2014. The selected signalized intersection was located on

Page 7 of 42



Draft

8

Jiangdong Road with Aoti Street. Jiangdong Road is a north-south eight-lane divided highway

serving as a major corridor in the west area of Nanjing city. The average annual daily traffic on

Jiangdong Road exceeded 30,000 vehicles per day in the year 2009 (Liu et al. 2013). Aoti Street

is a west-east minor street. Both traditional inside and unconventional outside left-turn lanes are

installed at the southbound and northbound approaches of the signalized intersection, while only

the traditional inside left-turn lanes are installed at the westbound and eastbound approaches of

the signalized intersection (see Figure 1). Protected left-turn phases are also provided to

accommodate the left-turn movements at the signalized intersections. The unconventional left-

turn lanes were installed at the intersection in 2005 with the aim of improving the mobility of

left-turn traffic flow. Figure 1 also shows the six left turn lanes numbered as well as examples of

potential conflicts. Field observation showed that the left-turn movements at the southbound and

westbound approaches lead to four main types of traffic conflicts related to the outside left-turn

lanes. The rear-end conflicts take place at both of the outside and inside left-turn lanes (Figure 1

(a)). The left-turn movements from the outside and inside left-turn lanes at the same approach can

result in sideswipe conflicts and merging conflicts (Figure 1 (b) and (c)). The left-turn

movements from the outside and inside left-turn lanes at the opposite approaches can lead to

head-on conflicts (Figure 1 (d)).

Video recording of the location was done for 2.5 hours on May 12, 2011. The video frame

rate is 30 frames/sec. The camera was placed on top of a roadside building to achieve adequate

viewing height. Field data collection was conducted during a weekday morning peak period

under sunny weather. The aim is to eliminate the impact of weather factor, such as rainy, and fog

and haze, on the traffic flow.

4. Computer Video Analysis

Page 8 of 42



Draft

9

The computer video analysis system used in the study was developed at the University of British

Columbia (Saunier and Sayed 2006; Saunier et al. 2010), which includes two modules as shown

in Figure 2. The computer video analysis system can automatically detect, classify, and track road

users and interpret their movements. The main steps of the analysis are summarized below.

Camera calibration: The outcome of the camera calibration is a mapping between the

three dimensional real word and the two dimensional image space (Figure 3). Details of the

mixed-feature camera calibration approach can be found in Ismail et al. (Ismail et al. 2013). The

accuracy of the estimated parameters was tested by use of a set of line segments of true length

estimated from the orthographic image. This set of observations was not used in the calibration.

The calibration error is represented by the discrepancy between the calculated and the annotated

segment lengths normalized by the length of each segment. The accuracy of the final estimates

was satisfactory, and no further error in safety analysis was attributed to inaccurate estimated

camera parameters.

Features tracking: In this step distinguishable features such as points or lines on the

moving objects are tracked (Figure 4 (a)). The automated video analysis relies on computer

algorithms to differentiate between features of road users and features that are part of the

environment. Features that remain stationary are assumed to belong to the environment are

discarded and not tracked. Features on moving objects are identified and tracked using an

implementation of Kanade-Lucas-Tomasi (KLT) feature tracker algorithm (Saunier and Sayed

2006). The KLT algorithm has the advantages of consistently tracking objects under different

temporal and light conditions.

Features grouping: Moving objects may have multiple features. Feature grouping is

carried out using cues such as spatial proximity and common motion of features to decide

common features of each unique road-user. Features with motion vectors differing by more than a

Page 9 of 42



Draft

10

specified threshold are assumed to not belong to the same road-user object and are not grouped,

regardless of their spatial proximity. Figure 4 (b) shows moving objects generated by grouping of

features. The tracking accuracy for road users was validated by Saunier and Sayed (Saunier and

Sayed 2006) and Ismail et al. (Ismail et al. 2010) with accuracy between 90% and 94.4% on

different sets of sequences. This accuracy is considered reliable and has little impact on the

accuracy of the calculation of conflict indicators.

Objects classification: The object trajectories hold features that reveal the structure of

the traffic movements and provide important clues to the characteristics of the road users.

Basically, the road user classification method is based on three indicators, namely, movement

patterns, maximum speed, and road user size. The cyclist has a special movement pattern which

is governed by its pedaling process compared to vehicles, leading to significant periodic

variations in the speed profile. Vehicle movement patterns are primarily composed of linear

segments corresponding to different speeds chosen throughout the trajectories. This oscillatory

behavior associated with cyclist while lacking in vehicles, provides a classification basis through

recognizing the class of the road-user (Zaki and Sayed 2013). Subsequently, the maximum speed

and road-user size are used as complimentary cues to enhance the classification result. The road

user classification has been investigated by Zaki and Sayed (Zaki and Sayed 2013) with accuracy

above 90%. Figure 4 (c) shows different objects of vehicle (yellow box) and bike (white box).

Event generation: The generation of interactions between objects can be performed

based on the trajectories extracted from the computer vision module. The trajectory of a moving

object is matched to individual prototypes (See Figure 4 (d) for illustrative prototypes) from the

full set of prototypes using a Longest Common Sub Sequence (LCSS) algorithm (Saunier et al.

2010).

Page 10 of 42



Draft

11

LCSS Algorithm: Let τ be a finite set of road user tracks {Ti |i ϵ 1,…,N(τ)}, with N(.) is a

measure on a finite set that returns the number of the set elements. Let each road user track Ti be

composed of a set of coordinate tuples such that Ti= {tik |k ϵ 1,…,N(Ti)} and each coordinate tuple

tik be defined as tik= (xik,yik). Two points tik and tjl to be matched if max{|xik-xjl|,|yik-yjl |} < ε, where

ε is some spatial proximity bound called matching distance. The LCSS of two road user tracks Ti

and Tj, LCSSε (Ti,Tj) of respective lengths m and n, is defined recursively as follows:

• 0�� = 0�� = 0, • 1 + ��(��(��, ��, ��ℎ�� !��"ℎ,

• max & ��(��, ��, �� '��, ��() , ��ℎ��*��

where Head(Ti)= {tik |k ϵ 1,… ,N(Ti)-1} and the definition is identical for all tracks other than i.

The LCSS algorithm compares the tracks against a set of templates (prototypes) of

expected road-user behavior at the given intersection. The computer vision system described

earlier has a built-in procedure to extract a set of common tracks of road-users. However, more

often the set of generated prototypes do not provide adequate representative of the road-user

tracks. This is primary depending on the footage length used in the prototype generation as well

as the distinct tracks found in the footage. An iterative procedure may be implemented to ensure

certain behavior coverage. An alternative procedure, would be synthesize prototypes to cover

certain behavior that deemed hard to extract from the footage. An algorithm is developed to

generate prototypes representing behavior at the designated area. An object will therefore be

matched with more than one prototype with a probability weighting determined from the LCSS

matching distance. The prototypes provide a set of predicted future positions with associated

probabilities of occurrence. Conflicts between road users can then be determined by evaluating if

Page 11 of 42



Draft

12

any of these future positions coincide spatially and temporally with other road users. Figure 4 (e)

shows an example of rear-end conflict event between two left-turn vehicles.

Conflict indicator filtering: This step is to analyze the severity of the identified conflict

events using conflict indicators. The TTC indicator is used in this study as a measure of conflict

severity. The TTC is defined as the time until a collision will occur if the two conflicting vehicles

were to continue on the same path at their current speed as shown in Figure 5 (a) and Figure 5 (b).

Therefore, TTC is continually calculated between conflicting road users, and thus a set of values

is returned for each conflict according to the conflicting vehicles instantaneous speeds. The

minimum TTC is then extracted from this set to represent the maximum severity of each

interaction (Sayed and Zein 1999). To reduce the noise influence on the TTC calculation, the

following steps were conducted. First, a low pass filtering is applied to smooth out the

trajectories and eliminate tracking noise. Second, the choice of the minimum TTC was based on

averaging over few frames, rather than just selecting a global minimal value. Typically, the most

severe value is used to represent the overall severity of a traffic event. In this study, only traffic

events with associated minimum TTC of fewer than 4 sec were considered for evaluation. This

threshold was suggested by Horst (Horst 1991) to distinguish between safe and uncomfortable

situations on the roads and it was supported by Farber (Farber 1991) and Osama et al. (Osama et

al. 2015).

A significant portion of the study focuses on validating the quality of the automated event

detection. First, video data were manually reviewed by an expert to annotate possible conflicts

between the road-users using the definition given in this paper and the US FHWA observer’s

guide (Parker and Zegeer 1989). While manual conflict analyses suffer from the shortfalls

previously described, the length of this video allowed for an extremely vigilant review. Each

conflict was described by the manual reviewer in detail so that it could later be compared to the

Page 12 of 42



Draft

13

automatically tracked events. Validation was performed on a subset of 50 events selected from

the interactions database and results showed a 95.6% accuracy of the automated TTC index

estimation. The scope of the validation was limited to a comparison between an event's minimum

TTC and a corresponding manually calculated TTC. Each conflict was also assigned a severity

rating by the reviewer. Though this measure was highly subjective, it was deemed to have

sufficient merit as the engineer had experience in such reviews. The results demonstrated the

accuracy of the automated TTC index estimation.

5. Summary of Findings

5.1 Overview of the whole intersection

5.1.1 Conflicts spatial distribution analysis

The conflict analysis includes identifying conflict frequency, severity and location (conflict

points). Five types of conflicts are considered of importance for the safety diagnosis of the

intersection: rear-end conflicts, merging conflicts, sideswipe conflicts, head-on conflicts and

crossing conflicts1. Traffic conflicts between vehicles and between vehicles and bikes

2 at the

whole intersection were automatically identified. In total, 5893 cars and 2462 bikes were tracked.

The density of traffic conflicts per unit area was also measured. Spatial distribution of the

vehicle-vehicle conflicts and vehicle-bike conflicts by heat mapping is shown in Figure 6. Figure

6 (a) shows that vehicle-vehicle conflicts covered the whole intersection but were concentrated

along Jiangdong road. This may be caused by the high traffic volume and the presence of two

left-turn lanes per approach on Jiangdong road. The highest conflict density is found at the inner

intersection caused by the left-turn movements, as well as the southbound exit resulting from the

1 A crossing conflict refers to conflicts between a through vehicle and a vehicle crossing from street on the left or right. 2 All types of bikes are considered in the analysis including motorcycles, e-bikes and bicycles.

Page 13 of 42



Draft

14

high volume of north-south direction. Figure 6 (b) shows that vehicle-bike conflicts take place

anywhere at the intersection. An observation from the video reveals that the violation behavior

of bikes in red light running, occasional reverse driving, waiting at violating positions, and bikes

moving in motor-vehicle lanes result in the vehicle-bike conflicts.

5.1.2 Conflicts frequency analysis

Table 1 shows a breakdown of the frequency (per hour) of vehicle-vehicle conflicts and vehicle-

bike conflicts according to each TTC value. It shows that a high percentage of conflicts have

TTC values fewer than 2 seconds for both the vehicle-vehicle conflicts and vehicle-bikes

conflicts, which is considered as high severity conflicts.

Figure 7 shows the frequency distributions of conflicts at the intersections for both

vehicle-vehicle conflicts and vehicle-bike conflicts. Rear-end, merging, sideswipe, head-on,

crossing, and total conflicts are displayed separately. The distributions are plotted over a range of

TTC values from 0 to 4 sec. The figures show that the frequency of vehicle-vehicle conflicts is

higher than that of vehicle-bike conflict for all types of conflicts. The only exception is with the

crossing conflict where the conflict frequency curves for vehicle-vehicle conflicts and vehicle-

bike conflicts overlap.

5.1.3 Conflicts severity analysis

To further evaluate the conflict severity, the minimum TTC of each event can be mapped to a

severity index using equation (1) (Saunier et al. 2010; Autey et al. 2012).

2

2

TTCexp

2PRTSI

= −

(1)

where SI is the severity index and PRT is the perception and braking reaction time, which is

assumed to be 2.5 sec (Autey et al. 2012). The severity index is a unit-less measure of severity

that ranges from 0 to 1, with 0 being uninterrupted passages.

Page 14 of 42



Draft

15

Table 2 shows a breakdown of the number of vehicle-vehicle conflicts and vehicle-bike

conflicts by severity. Figure 8 shows the severity distributions of conflicts at the intersections for

both vehicle-vehicle conflicts and vehicle-bike conflicts. Rear-end, merging, sideswipe, head-on,

crossing, and total conflicts are displayed separately. The conflicts frequency is plotted over a

range of severity values from 0 to 1 sec. The results show that considerable numbers of conflicts

have high severity index (SI < 0.4) for both vehicle-vehicle conflicts (43%) and vehicle-bike

conflicts (40%). Similar as the result of conflict frequency distribution, it is found that the

severity index of vehicle-vehicle conflicts is higher than that of vehicle-bike conflicts for all

types of conflict expect for the crossing conflict which has similar severity between the conflicts

of vehicle-vehicle and vehicle-bike.

5.2 Comparison of left-turn conflicts from different left-turn lanes

Due to the small sample size of the vehicle-bike conflicts for left-turn movements (12 events per

hour), the following sections focus on vehicle-vehicle conflicts. The conflict frequency between

left-turn lanes at southbound (SB) and northbound (NB) approaches is shown in table 3. It should

be noticed from table 3 that there is a considerable number of head-on conflicts caused by the

outside left-turn lane (L1-L3, L1-L4, and L2-L4, See Figure 1 for Lane annotations). As the

head-on conflicts are considered as the most severe conflicts, the result indicates that the outside

left-turn lanes can have serious safety impacts on the intersection.

The conflict frequency (per hour) and percentage of vehicles involved in conflicts of each

left-turn lane is shown in Figure 9. The percentage of vehicles involved in conflicts of each left-

turn lane is calculated using Equation (2).

Percentage of vehicles per lane involved in conf 1icts 00l i

i

C

V= × (2)

Page 15 of 42



Draft

16

where Ci represents the conflicts frequency (per hour) of left-turn lane i; Vi represent the hourly

traffic volume of left-turn lane i.

As shown in Figure 9, lane 2 (inside left-turn lane) is found to have the highest conflict

frequency, followed by lane 1 (outside left-turn lane). Lane 3 (inside left-turn lane) is found to

have the highest percentage of vehicles involved in conflicts, followed by lane 4 (outside left-turn

lane). However, approximately 35% of the conflicts at lane 2 are caused by the outside left-turn

lanes, and over 60% of the conflicts at lane 3 are caused by the outside left-turn lanes as shown in

table 4. The results suggest that the outside left-turn lane not only contributed to conflicts within

the lane itself, but also contributed to conflicts with the other left-turn lanes. The results also

show that the inside left-turn lanes at the eastbound (EB) and westbound (WB) approaches where

only inside left-turn lanes were installed have the smallest conflict rate compared to other left-

turn lanes at southbound and northbound approaches.

5.3 Comparison of left-turn conflicts from different approaches

To further illustrate the safety impacts of outside left-turn lane, the conflicts frequency and rate

between the four approaches were compared. As shown in Figure 10, the conflict frequency at SB

approach is maximum with a very high value of 98 conflicts per hour compared to other

approaches. This is caused by the high left-turn traffic volume from southbound approach.

However, NB approach is found to have the highest percentage of vehicles involved in conflicts

with the value of 75%. The low left-turn traffic volume from northbound approach contributes to

the high percentage of vehicles involved in conflicts. The lower conflicts frequency and

percentage of vehicle involved are found at EB approach and at WB approach. In addition, the

outside left-turn lane provides plenty of merging conflicts and sideswipe conflicts with the inside

left-turn lane at SB approach and NB approach.

Page 16 of 42



Draft

17

To further clarify the comparison, the left-turn conflicts at approaches with both outside

and inside left-turn lanes and at approaches with only inside left-turn lanes are aggregated and

compared. As shown in Figure 11, both the conflict frequency and percentage of vehicles

involved in conflicts at SB and NB approaches are significantly higher than that at EB and WB

approaches. The results indicate that approaches with outside left-turn lanes have considerably

more conflicts than approaches with only inside left-turn lanes at signalized intersections.

The frequency of conflicts, normalized to exposure, is plotted over a range of severity

values as shown in Figure 12. The results reveal a higher severity index at the approaches with

outside left-turn lanes. A statistical t-test was applied to identify if the difference between

severity indices at SB and NB approaches and at EB and WB approaches is statistically

significant. Result of the t-test shows that the severity index of conflicts related to outside left-

turn lanes is significantly higher than that related to inside left-turn lanes with a 90% level of

confidence (0.03 & 0.011, p = 0.064).

The spatial distribution heat maps for the traffic conflicts from different left-turn lanes are

shown in Figure 13. The outside left-turn lane approaches have higher conflict density compared

to the non-outside left-turn lane approaches. Specifically, the conflict density for L1 and L2 are

higher compared to other left-turn lanes. As mentioned above, the high conflict density for L2

includes conflicts between L2 and other outside left-turn lanes. Although conflict density for L4

is low, most of the conflicts are head-on conflicts. It should also be noted that conflict density for

L5 and L6 are significantly lower than the other left-turn lanes. These results confirm that outside

left-turn lanes contribute to higher conflict frequency and severity at signalized intersections.

6. Conclusion

Page 17 of 42



Draft

18

The research presented in this paper evaluated the safety effects of an unconventional outside

left-turn lane deign at signalized intersection. The evaluation was conducted using an automated

video-based traffic conflict technique. Several observations can be drawn from the above-

mentioned analysis. First, although the outside left-turn lane provided a higher speed of left-turn

vehicles (Liu et al. 2011), it leads to negative safety impacts on signalized intersections. The

results of the analysis showed that the approaches with outside left-turn lanes had considerable

conflicts compared to approaches without outside left-turn lanes. Second, the approaches with

outside left-turn lanes had significantly higher severity than approaches without outside left-turn

lanes. Third, the outside left-turn lane not only contributed to conflicts within the lane itself, but

also contributed to conflicts with the other left-turn lanes. As well, the outside left-turn lane leads

to plenty of head-on, merging, and sideswipe conflicts.

The findings of this study provide new insight into the unconventional outside left-turn

lanes. The results show that outside left-turn lanes increase the frequency and severity of traffic

conflicts inside the signalized intersections. However, it should be noted that the left-turning

vehicles do not need to make lane changes to the most inside lane to make left turns. As such, it

is expected that the outside left-turn lanes could reduce the conflicts at the upstream areas of the

intersection. It is recommended that the trade-off between the benefits at upstream and negative

safety impact at the intersection caused by the outside left-turn lanes be carefully considered

before recommending their installation. Several limitations need to be considered however in

future studies. First, the results of the present study are based on a short observational periods,

further study should be conducted using a long time video data. Second, a before-after study

should further strengthen the results presented in the study. Third, due to some drivers not being

familiar with the new design, some violation behaviors such as aggressive lane changing and

driving in wrong direction were observed. The violation and driving behavior analysis was an

Page 18 of 42



Draft

19

extending of the current research. Fourth, a safety diagnose of the outside left-turn lanes should

be conducted that extends to the upstream areas of signalized intersections. Further studies should

also be conducted to examine the travel speed at the outside and inside left-turn lanes.

Page 19 of 42



Draft

20

Acknowledgments

This research was sponsored by the National Natural Science Foundation of China (Grant No.

51322810), the Fundamental Research Funds for the Central Universities (Grant No. YBJJ1458),

and the Scientific Innovation Research of College Graduates in Jiangsu Province (Grant No.

KYLX_0174). The authors also would like to thank the graduate research assistant at the School

of Transportation at the Southeast University for their assistance in field data collection.

Page 20 of 42



Draft

21

Reference

Amundsen, F., and Hydén, C. 1977. First Workshop on Traffic Conflicts. Institute of

Transportation Economics. Oslo, Norway.

Autey, J., Sayed, T., and El Esawey, M. 2013. Operational performance comparison of four

unconventional intersection designs using micro-simulation. Journal of Advanced

Transportation. 47(5): 536-552. doi: 10.1002/atr.181.

Autey, J., Sayed, T., and Zaki, M.H. 2012. Safety evaluation of right-turn smart channels using

automated traffic conflict analysis. Accident Analysis and Prevention. 45: 120-130.

doi:10.1016/j.aap.2011.11.015.

Brown, G.R. 1994. Traffic conflicts for road user safety studies. Canadian Journal of Civil

Engineering. 21(1): 1-15.doi: 10.1139/l94-001.

El Esawey, M., and Sayed, T. 2013. Analysis of unconventional arterial intersection designs

(UAIDs): state-of-the-art methodologies and future research directions. Transportmetrica

A: Transport Science. 9(10): 860-895. doi:10.1080/18128602.2012.672344.

Parker Jr. M.R., and Zegeer, C.V. 1989. Traffic conflict techniques for safety and operations-

observers manual. FHWA-IP-88-027, U.S. Department of Transportation, Federal

Highway Administration, Washington, D.C.

Farber, B. 1991. Designing a distance warning system from the user point of view. APSIS report,

Glonn-Haslach: Institute fur Arbeitspsychologie and Interdisziplinare Systemforchung.

Goldblatt, R., Mier, F., and Friedman, J. 1994. Continuous flow intersections. ITE Journal. 64(7):

35-42.

Hayward, J. 1972. Near-miss determination through use of a scale of danger. Highway Research

Record. 384: 24-34.

Page 21 of 42



Draft

22

Horst, R.V.D. 1991. Time-to-collision as a cue for decision making in braking. Vision in

Vehicles III. 19-26.

Hummer, J.E. 1998. Unconventional left‐turn alternative for urban and suburban arterials-part

one. ITE Journal. 68(9): 26–29.

Ismail, K., Sayed, T., and Saunier, N. 2010. Automated analysis of pedestrian-vehicle conflicts: a

context for before-and-after studies. Transportation Research Record: Journal of the

Transportation Research Board. 2198: 52-64. doi: 10.3141/2198-07.

Ismail, K., Sayed, T., and Saunier, N. 2013. A methodology for precise camera calibration for

data collection applications in urban traffic scenes. Canadian Journal of Civil

Engineering. 40(1): 57-67. doi: 10.1139/cjce-2011-0456.

Laureshyn, A., Ardo, H., Svensson, A., and Jonsson, T. 2009. Application of automated video

analysis for behavioural studies: concept and experience. IET Intelligent Transport

Systems. 3(3): 345-357. doi: 10.1049/iet-its.2008.0077.

Liu, P., Lu, J.J., Zhou, H.G., and Sokolow, G. 2007. Operational effects of U-turns as alternatives

to direct left-turns. Journal of Transportation Engineering. 133(5): 327-334. doi:

10.1061/(ASCE)0733-947X(2007)133:5(327).

Liu, P., Wan, J.J., Wang, W., and Li, Z.B. 2011. Evaluating the impacts of unconventional

outside left-turn lane design on traffic operations at signalized intersections.

Transportation Research Record: Journal of the Transportation Research Board. 2257: 62-

70. doi: 10.3141/2257-07.

Liu, P., Xu, C.C., Wang, W., and Wan, J.J. 2013. Identifying factors affecting drivers' selection

of unconventional outside left-turn lanes at signallised intersections. IET Intelligent

Transport Systems. 7(4): 396-403. doi: 10.1049/iet-its.2011.0229.

Page 22 of 42



Draft

23

Osama, A., Sayed, T., Zaki, M.H., and Shaaban, K. 2015. Automated approach for a

comprehensive safety assessment of roundabouts. Transportation Research Board 94th

Annual Meeting, Washington, D.C.

Perkins, S., and Harris, J. 1968. Traffic conflict characteristics: accident potential. Highway

Research Record. 225: 35-43.

Saunier, N., and Sayed, T. 2006. A feature-based tracking algorithm for vehicles in intersections.

The 3rd Canadian Conference on Computer and Robot Vision, Quebec.

Saunier, N., Sayed, T., and Ismail, K. 2010. Large scale automated analysis of vehicle

interactions and collisions. Transportation Research Record: Journal of the Transportation

Research Board. 2147: 42-50. doi: 10.3141/2147-06.

Sayed, T., and Zein, S. 1999. Traffic conflict standards for intersections. Transportation Planning

and Technology. 22(4): 309-323. doi: 10.1080/03081069908717634.

Sayed, T., Brown, G., and Navin, F. 1994. Simulation of traffic conflicts at unsignalized

intersections with TSC-Sim. Accident Analysis and Prevention. 26(5): 593-607. doi:

10.1016/0001-4575(94)90021-3.

Sayed, T., Ismail, K., Zaki, M.H., and Autey, J. 2012. Feasibility of computer vision-based safety

evaluations: case study of a signalized right-turn safety treatment. Transportation

Research Record: Journal of the Transportation Research Board. 2280: 18-27. doi:

10.3141/2280-03

Sayed, T., Zaki, M.H., and Autey, J. 2013. Automated safety diagnosis of vehicle–bicycle

interactions using computer vision analysis. Safety Science. 59: 163-172. doi:

10.1016/j.ssci.2013.05.009.

Page 23 of 42



Draft

24

Tabernero, V., and Sayed, T. 2006. Upstream signalized crossover intersection: an

unconventional intersection scheme. Journal of Transportation Engineering. 132(11): 907-

911. doi: 10.1061/(ASCE)0733-947X(2006)132:11(907).

Tageldin, A., Sayed, T., Zaki, M.H., and Azab, M. 2014. A safety evaluation of an adaptive

traffic signal control system using computer Vision. Advances in Transportation Studies.

2: 83-96. doi: 10.4399/97888548735379.

Tarko, A., Davis, G., Saunier, N., Sayed, T., and Washington, S. 2009. White paper: surrogate

measures of safety. Transportation Research Board, National Research Council,

Washington, D.C.

Zaki, M.H., and Sayed, T. 2013. A framework for automated road-users classification using

movement trajectories. Transportation Research Part C: Emerging Technologies. 33: 50-

73. doi: 10.1016/j.trc.2013.04.007.

Zaki, M., Sayed, T., and Cheung, A. 2013. Computer vision techniques for the automated

collection of cyclist data. Transportation Research Record: Journal of the Transportation

Research Board. 2387: 10-19. doi: 10.3141/2387-02.

Zaki, M.H., Sayed, T., Tageldin, A., and Hussein, M. 2013. Application of computer vision to

diagnosis of pedestrian safety issues. Transportation Research Record: Journal of the

Transportation Research Board. 2393: 75-84. doi: 10.3141/2393-09.

Page 24 of 42



Draft

25

TABLES

Table 1 Vehicle to vehicle and vehicle to bike conflict frequency per hour

Conflicts Vehicle to vehicle Vehicle to bike

TTC range # Event # Event

0-1s 123 17

1-2s 169 20

2-3s 142 13

3-4s 89 10

Total 523 60

Page 25 of 42



Draft

26

Table 2 Vehicle to vehicle and vehicle to bike conflict severity distribution per hour

Conflicts Vehicle to vehicle Vehicle to bike

SI range # Event # Event

0-0.2 126 14

0.2-0.4 101 9

0.4-0.6 86 11

0.6-0.8 87 9

0.8-1.0 123 17

Total 523 60

Page 26 of 42



Draft

27

Table 3 Conflict frequency per hour between different left-turn lanes

Lane No. L1 L2 L3 L4

L1 16 15 13 4

L2 15 29 4 2

L3 13 4 5 1

L4 4 2 1 0

Page 27 of 42



Draft

28

Table 4 Conflict frequency at left-turn lanes and proportion of conflicts related to outside left-

turn lane

Lane No. Frequency Proportion of conflicts related to

outside left-turn lane

L1 48 100%

L2 50 35.3%

L3 23 62.5%

L4 7 100%

L5 11 0

L6 3 0

Page 28 of 42



Draft

29

FIGURE CAPTIONS

Figure 1 Layout of the analyzed intersection and possible conflicts types for left-turn

movements

Figure 2 Video analysis block diagram

Figure 3 Camera calibration results

Figure 4 Demonstration of the video analysis process

Figure 5 TTC and speed of objects for vehicle to vehicle and vehicle to bike conflicts

Figure 6 Spatial distribution heat maps for different types of conflicts at the intersection

Figure 7 Different types of conflicts frequency distribution

Figure 8 Conflicts severity distribution varied different types

Figure 9 Conflict frequency and percentage of vehicles involved in conflicts of different left-

turn lanes

Figure 10 Conflicts frequency and percentage of vehicles involved in conflicts at four approaches

Figure 11 Conflicts frequency and percentage of vehicles involved in conflicts at different

approaches

Figure 12 Conflicts severity distribution for different approaches

Figure 13 Spatial distribution heat maps for left-turn conflicts from different left-turn lanes

Page 29 of 42



Draft

Figure 1 Layout of the analyzed intersection and possible conflicts types for left-turn movements 152x131mm (300 x 300 DPI)

Page 30 of 42



Draft

Figure 2 Video analysis block diagram

37x8mm (300 x 300 DPI)

Page 31 of 42



Draft

Figure 3 Camera calibration results

179x154mm (300 x 300 DPI)

Page 32 of 42



Draft

Figure 4 Demonstration of the video analysis process 266x343mm (300 x 300 DPI)

Page 33 of 42



Draft

Figure 5 TTC and speed of objects for vehicle to vehicle and vehicle to bike conflicts 194x203mm (300 x 300 DPI)

Page 34 of 42



Draft

Figure 6 Spatial distribution heat maps for different types of conflicts at the intersection 81x32mm (300 x 300 DPI)

Page 35 of 42



Draft

Figure 7 Different types of conflicts frequency distribution 181x172mm (300 x 300 DPI)

Page 36 of 42



Draft

Figure 8 Conflicts severity distribution varied different types 181x170mm (300 x 300 DPI)

Page 37 of 42



Draft

Figure 9 Conflict frequency and percentage of vehicles involved in conflicts of different left-turn lanes 100x59mm (300 x 300 DPI)

Page 38 of 42



Draft

Figure 10 Conflicts frequency and percentage of vehicles involved in conflicts at four approaches 100x59mm (300 x 300 DPI)

Page 39 of 42



Draft

Figure 11 Conflicts frequency and percentage of vehicles involved in conflicts at different approaches 100x59mm (300 x 300 DPI)

Page 40 of 42



Draft

Figure 12 Conflicts severity distribution for different approaches 100x59mm (300 x 300 DPI)

Page 41 of 42



Draft

Figure 13 Spatial distribution heat maps for left-turn conflicts from different left-turn lanes 239x294mm (300 x 300 DPI)

Page 42 of 42



DraftDraft 1 Safety Evaluation of Unconventional Outside Left-Turn Lane Using Automated Traffic...

Documents

Transcript of DraftDraft 1 Safety Evaluation of Unconventional Outside Left-Turn Lane Using Automated Traffic...